Dim Moving Target Algorithm Research Based On Sparse Bayesian Learning

The infrared dim small target detection and tracking technology under complex background is a key technology in infrared guidance and air defense early warning.Because of the noise of infrared imaging system and the distance from detector to the target,the target is usually shown as a small and weak target with low luminance value and few pixels.It is difficult to detect targets because of the strong noise.Therefore,it is of theoretical and practical significance to study an infrared dim small target detection technology with low SNR.Sparse Bayesian method combines structural prior information with sample information,use probability to express the uncertainty of all forms,Learning the state of signals through Bayes theorem,Making full use of the spatial structure and temporal structure of signal sparse representation coefficients,with excellent parameterized performance.Although Sparse Bayesian theory is still in the stage of development and exploration,it shows great potential in the field of signal processing.Based on the above background,the infrared dim small moving target method in the image sequence is deeply researched using Sparse Bayesian theory in the thesis.The main research results of the full thesis are as follows:(1)The energy characteristics and image scene characteristics of infrared dim small target images are analyzed.The results show that the target energy is outstanding,and the background clutter is obvious,resulting in obvious clutter.(2)The time and space sparse characteristics of the target images are analyzed.The experimental results show that there is a strong correlation between the sparse representation of the image in the time domain and the main feature of the target image after the clustering of the space atom.(3)The sparse decomposition coefficient of continuous frame images is fitted and tested by statistical distribution,and the best fitting effect of Gaussian function distribution model is constructed by comparing the fitting effect of different kernel functions.(4)An infrared dim small target detection method based on sparse Bias is proposed.The spatial structure and time sequence structure of the presequence frame coefficients are fully studied.The distribution functions of the sparse decomposition coefficients are obtained for the image blocks in the infrared images.By using the Bias minimum error rate criterion,the image blocks are sparse by the prior object and the background image block.The distribution function of decomposition coefficient is compared to determine whether the test block is the target image block,and the target detection is completed.The experimental results show that the sparse Bayesian theory will be introduced into the infrared small and weak target detection field.The prior information of the presequence frame of the image signal can be applied to the target detection of the subsequent frames,which overcomes the defect that the traditional spatial dictionary can only describe the target morphology,and is more conducive to the detection of the target.